Abstract
Purpose
To evaluate the benefits of combining the Periodically Rotated Overlapping ParallEL Lines with Enhanced Reconstruction (PROPELLER) acquisition technique and deep learning—based reconstruction (DLR) for fat-suppressed T2-weighted imaging (Fs-T2WI) and diffusion-weighted imaging (DWI) in head and neck MRI.
Materials and methods
This retrospective study included 34 patients who underwent 3.0-T head and neck MRI. Imaging protocols comprised PROPELLER-based Fs-T2WI and DWI, which were compared against conventional multiplanar fast spin-echo Fs-T2WI and single-shot echo-planar imaging DWI. All sequences were reconstructed using a DLR algorithm. Two radiologists independently performed qualitative assessment, evaluating overall image quality, lesion conspicuity, anatomical delineation, and artifact severity using a 5-point Likert scale. The quantitative assessment involved measurements of the signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) in the lesions, adjacent muscle, parotid glands, and submandibular glands. Interobserver agreement was determined using weighted kappa statistics, and Wilcoxon signed-rank test was used for statistical comparisons.
Results
The PROPELLER sequences exhibited significantly higher qualitative scores for all evaluated parameters in both Fs-T2WI and DWI compared to the conventional sequences (p < 0.001). The interobserver agreement for the Fs-T2WI was moderate (0.47–0.53), and that for DWI was good (0.76–0.83). The quantitative analysis further demonstrated significantly higher SNRs and lesion-to-muscle CNRs with the PROPELLER sequences (p < 0.001).
Conclusion
The combination of PROPELLER acquisition and DLR significantly improves the image quality and lesion conspicuity in head and neck MRI. This approach effectively suppresses artifacts and improves quantitative image metrics, thereby positioning it as a reliable imaging strategy for routine clinical assessments of head and neck lesions.
Keywords: Head and neck, PROPELLER, T2WI, DWI, Deep-learning reconstruction, Image quality
Highlights
-
•
PROPELLER acquisition + DLR significantly improves head and neck MRI image quality.
-
•
PROPELLER acquisition + DLR improves lesion conspicuity and anatomical delineation.
-
•
DLR-PROPELLER improves the SNR, CNR, and lesion conspicuity vs. conventional MRI.
-
•
DLR-PROPELLER provides superior artifact suppression in Fs-T2WI and DWI sequences.
-
•
DLR-PROPELLER is a reliable option for routine clinical head and neck MRI.
1. Introduction
Magnetic resonance imaging (MRI) is widely used for anatomical assessments, lesion detection, differential diagnoses, and evaluations of the lesion extent in the head and neck region. Specifically, single fast spin-echo (FSE)-based T2-weighted imaging (T2WI) is commonly applied for structural delineation and lesion identification [1], [2], [3], [4]. In routine clinical practice, fat-suppressed (Fs) techniques are often combined with T2WI to suppress the high signal intensity from fat surrounding a lesion, thereby improving the lesion conspicuity. However, the complex anatomy of the head and neck region, characterized by numerous small structures, often necessitates high-spatial-resolution images. Such images inherently reduce the signal-to-noise ratio (SNR), potentially limiting the diagnostic quality. Although increasing the number of signal averages can compensate for this SNR loss, these adjustments prolong the acquisition times and consequently increase the risk of motion artifacts, typically caused by the examinee's swallowing, tongue movement, and/or respiration [5].
Diffusion-weighted imaging (DWI) is also widely applied for indirectly visualizing the diffusion properties of water molecules in vivo, providing clinically useful information by incorporating diffusion phenomena into the image contrast [6], [7], [8]. The most widely used readout scheme for acquiring DWI is single-shot echo-planar imaging (EPI), which offers rapid data acquisition and images with a relatively high SNR. However, EPI-based sequences are highly sensitive to local static magnetic field inhomogeneities, which are typically caused by air or metallic substances in the oral cavity, leading to image distortion and susceptibility artifacts. In addition, similar to T2WI, the image quality obtained with DWI is susceptible to motion artifacts. These factors can hinder accurate lesion evaluations [9], [10], [11].
The Periodically Rotated Overlapping ParallEL Lines with Enhanced Reconstruction (PROPELLER) readout method was developed to reduce motion artifacts while maintaining a high SNR [12]. In PROPELLER imaging, data are acquired as concentric strips of parallel lines rotating around the center of k-space. This acquisition strategy enables the correction of spatial inconsistencies (thereby reducing motion artifacts) and maintains a high SNR in the k-space center due to the overlap of multiple signal acquisitions [13], [14], [15], [16], [17]. However, the PROPELLER technique is known to require longer scan times. Moreover, image blurring and a relatively low SNR are often problematic in DWI acquisitions due to its long echo spacing and larger number of echo trains compared to conventional EPI-based DWI [18].
The recent development of deep learning-based reconstruction (DLR) techniques enables effective noise reduction that increases or maintains the SNR while shortening the scan time compared to conventional reconstruction methods [19], [20], [21]. Several studies have reported the utility of MRI protocols incorporating DLR in the head and neck region, demonstrating improved image quality [22], [23], [24], [25]. Particularly for the head and neck region, the combination of PROPELLER acquisition and DLR is expected to be highly effective for reducing the motion artifacts that are commonly observed in this region of the body, as PROPELLER provides inherent motion correction and DLR offers excellent noise reduction. For DWI, PROPELLER combined with DLR is also considered useful for achieving reduced image distortion and susceptibility artifacts, due to its FSE-based readout design, which inherently mitigates these issues compared to conventional EPI-based sequences.
We conducted the present study to evaluate the image quality and lesion visibility in head and neck MRI using PROPELLER sequences reconstructed with a DLR algorithm.
2. Materials and methods
This retrospective study was approved by our institutional review board, with a waiver of informed consent.
2.1. Patient population
The cases of 65 patients who presented to our institution for a head and neck assessment and underwent MRI on the system described below during the 12-month period from January to December 2024 were retrospectively screened. This study period corresponded to a transitional phase before the full clinical implementation of the PROPELLER-based MRI acquisition with the DLR protocol at our institution. During this period, PROPELLER-based sequences were acquired in clinical cases in addition to the conventional protocol. Patient inclusion was not prospectively planned; rather, the data obtained during this transitional phase were analyzed retrospectively. The inclusion criteria were: (1) availability of PROPELLER-based Fs-T2WI or DWI sequences, (2) availability of both PROPELLER and conventional image datasets, and (3) the use of a specific fat-suppression technique (described below). Cases were excluded if DWI was not acquired using a standard imaging protocol for matrix size and slice thickness. Ultimately, the cases of 34 patients fulfilled all of the inclusion criteria (30 for Fs-T2WI and 21 for DWI).
2.2. Imaging parameters
All scans were performed on a 3.0-T MRI system (Discovery MR750w; GE Healthcare, Waukesha, WI, USA) equipped with a GEM head-and-neck coil. The scan range was set in order to center the lesion within the head and neck region, extending from the skull base to the larynx. Each image acquisition included PROPELLER-based Fs-T2WI, conventional Cartesian FSE Fs-T2WI, PROPELLER-DWI, and conventional single-shot EPI-DWI sequences. For the Fs-T2WI, spectral attenuated inversion recovery (SPAIR) was used for fat suppression in both the PROPELLER and conventional acquisitions. For the DWI, chemical-shift–selective fat suppression was used for PROPELLER-DWI, while inversion-recovery–based fat suppression was employed for EPI-DWI to address the common inadequacy of fat suppression in the head and neck region. Representative imaging parameters are summarized in Table 1.
Table 1.
The acquisition parameters investigated for the head and neck MRI results in this study.
| Parameter | Conventional MRI | PROPELLER MRI | ||
|---|---|---|---|---|
| Fs-T2WI | DWI | Fs-T2WI | DWI | |
| TR (ms) | 4735 | 4000 | 9109 | 4000 |
| TE (ms) | 85.0 | 80.0 | 85.0 | 80.0 |
| b-value | - | 1000 | - | 1000 |
| Echo train lengths | 21 | 1 | 24 | 20 |
| Flip angle (°) | 142 | 90 | 130 | 110 |
| Field of view (mm) | 220 | 220 | 220 | 220 |
| Section thickness/gap (mm) | 3.0/0.9 | 3.0/0.9 | 3.0/0.9 | 3.0/0.9 |
| Acquisition matrix | 512 × 512 | 128 × 128 | 512 × 512 | 128 × 128 |
| Reconstructed matrix | 512 × 512 | 256 × 256 | 1024 × 1024 | 256 × 256 |
| Number of Excitation | 1.0 | 9.0 | 1.5 | 2.0 |
| Number of slices | 27 | 27 | 27 | 27 |
| Noise reduction factor (%) | 75 | 75 | 75 | 75 |
| Acquisition time (min:s) | 3:19 | 4:06 | 3:48 | 4:16 |
DWI: Diffusion-weighted imaging; Fs-T2WI: fat-suppressed T2-weighted image; PROPELLER: periodically rotated overlapping parallel lines with enhanced reconstruction; TE: echo time; TR: repetition time.
For each acquisition scheme, images were reconstructed using a two-dimensional DL-based algorithm (AIR™ Recon DL, GE Healthcare) [26]. This DLR method provides noise reduction, truncation artifact suppression, and edge sharpening. The DLR pipeline processes raw k-space data with a deep convolutional neural network comprising approx. 4.4 million trainable parameters and 10,000 kernels. The algorithm is widely compatible with various pulse sequences, contrast settings, field strengths, and coil configurations, and it has been extended to PROPELLER imaging. Noise-reduction levels of 25%, 50%, and 75% are available; the 75% level was applied in this study.
2.3. Image analysis: Qualitative assessment
A blinded qualitative review was performed by two board-certified radiologists specializing in head and neck imaging, with 8 and 16 years of respective experience. The assessed the results of the axial Fs-T2WI and DWI obtained using PROPELLER and conventional sequences, based on four criteria: (i) global image quality, (ii) lesion depiction, (iii) delineation of anatomical details, and (iv) artifact severity. Each parameter was rated on a 5-point Likert scale, in which 1 point indicated 'very poor (non-diagnostic),' 2 points = 'poor (diagnostically usable),' 3 points = 'moderate (acceptable for diagnosis),' 4 points = 'good (minor diagnostic limitations),' and 5 points = 'excellent (virtually no diagnostic limitations).'
2.4. Image analysis: quantitative assessment
The quantitative evaluation involved the measurement of signal intensities within square regions of interest (ROIs) placed on axial sections encompassing the lesion, adjacent muscle (pterygoid or sternocleidomastoid), the parotid gland, and the submandibular gland. Three slices were selected for each target per patient. For the lesion and muscle, three consecutive slices were selected in the head-to-foot direction to maximize the lesion area, provided that the same muscle was consistently visible across all selected slices. For the parotid and submandibular glands, three contiguous slices encompassing the largest glandular area were selected. To ensure accurate measurements, intravascular signal, prominent noise, and artifacts were avoided. All of the ROIs were initially placed on PROPELLER sequences and then copied to the corresponding anatomical locations on conventional sequences. If gross noise (which is likely to introduce measurement bias) was present within an ROI, the ROI's position was adjusted, referencing both the PROPELLER and conventional images, in order to minimize noise inclusion. The SNR was calculated as the mean signal intensity within the ROI divided by its standard deviation (SD). The CNR was calculated as the difference between the mean signal intensity of the lesion ROI and that of the muscle ROI, divided by the SD of the muscle ROI [27]. All quantitative measurements were performed by a radiologist with 4 years of experience in head and neck imaging.
2.5. Statistical analyses
The degree of interobserver agreement in the qualitative evaluation was quantified using the weighted kappa coefficient, with agreement categorized as follows: 0.00–0.20, poor; 0.21–0.40, fair; 0.41–0.60, moderate; 0.61–0.80, good; and 0.81–1.00, excellent. The comparisons of qualitative scores, as well as the SNR and CNR values, were performed using the Wilcoxon signed-rank test. A threshold of p < 0.05 was used to define statistical significance. All statistical analyses were carried out using R (ver. 4.2.2).
3. Results
The study population consisted of 34 patients (20 men, 14 women; median age, 68 years; range, 2–85 years). The primary lesions identified on MRI included: neoplastic lesions (tongue carcinoma, n = 5; paraganglioma, n = 4; oropharyngeal carcinoma, n = 3; hypopharyngeal carcinoma, n = 3; schwannoma, n = 2; lymphangioma, n = 2; inverted papilloma, n = 2; nasopharyngeal carcinoma, n = 1; maxillary sinus carcinoma, n = 1; and liposarcoma, n = 1), inflammatory lesions (rhinosinusitis, n = 7; mastoiditis, n = 1; and pharyngeal edema, n = 1), and one case with no detectable lesion. T2WI data were available for 30 patients, allowing the evaluation of 29 lesions, 54 parotid glands, and 41 submandibular glands. Similarly, DWI data were available for 21 patients, allowing the evaluation of 20 lesions, 42 parotid glands, and 29 submandibular glands. Representative Fs-T2WI images are shown in Fig. 1, Fig. 2, and representative DWI images are shown in Fig. 3.
Fig. 1.
Representative Fs-T2WI images of a 51-year-old woman with a paraganglioma in the right parapharyngeal space. On conventional Fs-T2WI (A), motion artifacts were observed throughout the image, resulting in an unclear delineation of the lesion and a poor depiction of its internal characteristics (arrow), even with deep learning-based reconstruction (DLR). The salivary glands and surrounding muscle were also poorly visualized (arrowhead). In contrast, PROPELLER-based Fs-T2WI (B) substantially reduced motion artifacts, allowing a clearer depiction of the lesion, including its margins and internal characteristics (arrow). The surrounding anatomical structures and adjacent muscle were also visualized with superior overall image quality and minimal motion artifacts (arrowhead). Fs-T2WI: fat-suppressed T2-weighted imaging, PROPELLER: Periodically Rotated Overlapping ParallEL Lines with Enhanced Reconstruction.
Fig. 2.
Representative Fs-T2WI images of an 80-year-old woman with a facial nerve schwannoma in the left parotid space. On conventional Fs-T2WI (A), motion artifacts reduced the lesion's image quality (arrow) and surrounding structures, particularly in the oral cavity (arrowhead). In contrast, PROPELLER-based Fs-T2WI (B) reduced motion artifacts and improved the visualization of the lesion margins and internal structure (arrow) while reducing motion-related image degradation in the oral cavity (arrowhead).
Fig. 3.
Representative case: EPI- and PROPELLER-DWI. A 64-year-old man with a retention cyst in the nasal cavity. On EPI-DWI (A), severe structural distortion was observed. Moderate-to-severe susceptibility artifacts were also present, and the lesion (arrow) was difficult to identify because of marked image distortion and susceptibility effects. In contrast, PROPELLER-DWI (B) reduced the image distortion and susceptibility artifacts while preserving the signal-to-noise ratio (SNR). DWI: diffusion-weighted imaging, EPI: echo-planar imaging.
The qualitative scoring demonstrated that the use of the PROPELLER acquisition technique significantly outperformed the conventional sequences across all of the assessed parameters (overall image quality, lesion depiction, anatomical visualization, and artifact severity) for both Fs-T2WI and DWI (p < 0.001). The weighted kappa values indicated moderate agreement for Fs-T2WI (0.47–0.53) and good agreement for DWI (0.76–0.83) between the two readers. These results are summarized in Table 2.
Table 2.
The results of the qualitative evaluation of head and neck MRI by the two radiologists.
| Reader 1 |
Reader 2 |
Kappa-score | ||||||
|---|---|---|---|---|---|---|---|---|
| Conventional | PROPELLER | p-value | Conventional | PROPELLER | p-value | |||
| Fs-T2WI | Overall image quality | 4.03 ± 0.81 | 4.80 ± 0.48 | < 0.001 | 2.73 ± 0.78 | 4.40 ± 0.56 | < 0.001 | 0.502 |
| Lesion conspicuity | 4.00 ± 0.89 | 4.72 ± 0.59 | < 0.001 | 2.97 ± 0.78 | 4.52 ± 0.57 | < 0.001 | 0.528 | |
| Visualizations of anatomical structures | 4.07 ± 0.78 | 4.80 ± 0.48 | < 0.001 | 2.77 ± 0.77 | 4.40 ± 0.56 | < 0.001 | 0.474 | |
| Degree of artifacts | 3.73 ± 0.83 | 4.40 ± 0.67 | < 0.001 | 2.63 ± 0.76 | 4.10 ± 0.48 | < 0.001 | 0.488 | |
| DWI | Overall image quality | 2.05 ± 0.80 | 3.95 ± 0.80 | < 0.001 | 1.81 ± 0.68 | 3.90 ± 0.77 | < 0.001 | 0.833 |
| Lesion conspicuity | 2.40 ± 1.10 | 3.75 ± 0.71 | < 0.001 | 2.45 ± 0.76 | 4.05 ± 0.94 | < 0.001 | 0.801 | |
| Visualizations of anatomical structures | 1.62 ± 0.92 | 3.52 ± 0.68 | < 0.001 | 1.62 ± 0.67 | 3.71 ± 0.78 | < 0.001 | 0.758 | |
| Degree of artifacts | 1.90 ± 0.77 | 3.48 ± 0.75 | < 0.001 | 1.57 ± 0.68 | 3.81 ± 0.75 | < 0.001 | 0.770 | |
Data are mean ± standard deviation. DWI: Diffusion-weighted imaging; Fs-T2WI: fat-suppressed T2-weighted image; PROPELLER: periodically rotated overlapping parallel lines with enhanced reconstruction.
Quantitatively, the PROPELLER sequences demonstrated significantly higher SNR values (for the lesions, muscle, parotid gland, and submandibular gland) and CNR values than the conventional sequences in both Fs-T2WI and DWI (p < 0.001). These quantitative results are summarized in Table 3.
Table 3.
The quantitative SNR and CNR results for lesions, muscle, parotid glands, and submandibular glands.
| Conventional | PROPELLER | p-value | |||
|---|---|---|---|---|---|
| Fs-T2WI | SNR | Lesion | 25.90 ± 21.18 | 40.37 ± 39.20 | < 0.001 |
| Muscle | 8.05 ± 2.50 | 11.53 ± 4.16 | < 0.001 | ||
| Parotid gland | 9.16 ± 2.65 | 11.91 ± 3.70 | < 0.001 | ||
| submandibular gland | 14.42 ± 4.11 | 18.87 ± 4.48 | < 0.001 | ||
| CNR | Lesion to adjacent muscle | 37.28 ± 22.57 | 49.81 ± 31.69 | < 0.001 | |
| DWI | SNR | Lesion | 9.95 ± 7.48 | 19.13 ± 13.35 | < 0.001 |
| Muscle | 5.35 ± 1.08 | 6.69 ± 2.33 | < 0.001 | ||
| Parotid gland | 7.73 ± 3.04 | 9.40 ± 4.78 | < 0.001 | ||
| submandibular gland | 7.69 ± 2.47 | 9.76 ± 3.75 | < 0.001 | ||
| CNR | Lesion to adjacent muscle | 21.87 ± 26.14 | 38.83 ± 35.04 | < 0.001 |
Data are mean ± standard deviation. CNR: contrast-to-noise
ratio; DWI: Diffusion-weighted imaging; Fs-T2WI: fat-suppressed T2-weighted image; PROPELLER: periodically rotated overlapping parallel lines with enhanced reconstruction; SNR: signal to noise ratio.
4. Discussion
We evaluated the effectiveness of the PROPELLER technique combined with a deep learning-based image reconstruction algorithm for evaluations of the head and neck region. Compared to the conventional sequences reconstructed with the same deep learning-based algorithm, the DLR-PROPELLER combination demonstrated improved image quality in both the quantitative and qualitative assessments. Our search of the relevant literature identified few studies that examined DLR-PROPELLER Fs-T2WI and DWI, and the present study thus represents the first clinical investigation focusing specifically on head and neck imaging using this technique. The consistent improvement in image quality that we observed across various anatomical sites and lesion types suggests that this approach provides high-quality images that are suitable for comprehensive routine clinical evaluations and could facilitate more accurate image interpretations across a wide range of clinical tasks, including lesion detection and classification, cancer staging, and assessments of treatment responses.
The PROPELLER technique is an established imaging acquisition method for reducing motion artifacts, particularly in body regions that are prone to motion such as the neck, lungs, and abdomen [12], [13], [14], [17], [28], [29], [30], [31]. Motion artifacts caused by swallowing or respiration are especially problematic in clinical head and neck MRI [15], [16], [18]. The PROPELLER method involves acquiring strip-shaped data, termed "blades" (arranged like a propeller) that are repeatedly rotated around the center of k-space. The central region of k-space, which contains the most critical information for image contrast, is sampled multiple times in this technique. When patient motion affects the image, the motion-affected blades alone can be corrected or, in some cases, excluded from the reconstruction. This suppresses motion artifacts without significantly degrading the overall image quality [12], [30], [32], [33]. Concurrently, MRI sequences combined with DLR have been applied for imaging various organs including the brain, head and neck, and prostate [20], [21], [22], [23], [24], [25]. However, DLR-based noise reduction may not always be sufficient to eliminate the large-scale artifacts caused by patient bulk motion. Moreover, such artifacts may be misinterpreted by the reconstruction algorithm as true signal and can become more conspicuous after processing. Thus, conventional MRI sequences combined with DLR have inherent limitations in artifact reduction [26]. High-quality images can be obtained by applying DLR techniques to PROPELLER-acquired images, which already have substantially reduced motion artifacts, thereby improving the SNR. Several studies have combined the PROPELLER technique with DLR for imaging the body, cervical spine, and shoulder, demonstrating improvements in SNR values and image sharpness [28], [29], [34], [35]. In the present study, both the quantitative and qualitative evaluations yielded results consistent with these earlier investigations. The head and neck region in particular appears to be one of the most suitable anatomical areas for applying the combined DLR and PROPELLER sequence, likely due to the aforementioned reasons.
Single-shot EPI is the most commonly used conventional readout method for DWI. However, because EPI acquires k-space data from a single radiofrequency excitation using a long echo train, phase errors caused by magnetic field inhomogeneities tend to accumulate. This leads to substantial image distortion and susceptibility artifacts, particularly in the head and neck region, including the paranasal sinuses, temporal bone, oral cavity, pharynx, and larynx [18]. By contrast, the PROPELLER method is typically implemented as an FSE-based sequence in which phase errors induced by magnetic field inhomogeneities are largely refocused by 180° refocusing pulses, thereby reducing the image distortion and susceptibility artifacts relative to EPI [15], [17], [33], [36]. Nevertheless, DWI generally exhibits lower SNR values compared to T2WI, due to the signal attenuation associated with diffusion weighting. In addition, acquisition schemes that use long echo train lengths and prolonged echo spacing in the PROPELLER readout can lead to further SNR degradation caused by cumulative T2 decay [15]. To address this, the combination of PROPELLER acquisition and DLR can compensate for PROPELLER's associated SNR loss, ultimately facilitating the reconstruction of high-quality, high-resolution DWI [22], [23]. Our present analyses demonstrated that DLR-based PROPELLER-DWI effectively suppressed both motion and susceptibility artifacts while preserving high SNRs, resulting in superior image quality compared to the conventional approaches. Specifically, the PROPELLER-DWI provided excellent depiction of both lesions and normal anatomical structures, as confirmed visually.
Regarding the interobserver agreement obtained in this study, the weighted kappa values were lower for Fs-T2WI (moderate) than for DWI (good). This may reflect the fact that both PROPELLER and conventional Fs-T2WI were FSE-based, resulting in less pronounced image-quality differences than those observed for DWI, although the differences were still significant. In contrast, the larger difference in the readout scheme between PROPELLER-DWI and conventional EPI-DWI may have made the image-quality differences more conspicuous and easier to assess consistently.
4.1. Study limitations
Several limitations of this study should be acknowledged. (1) The sample size was limited (n = 34), and the study was conducted at a single institution; our findings should thus be considered preliminary. (2) The lesion types were heterogeneous, and the cohorts analyzed for Fs-T2WI and for DWI were not identical, resulting in smaller and unequal sample sizes for each sequence-specific analysis. The generalizability of these findings may therefore be limited, and further validation in larger, multicenter studies with more homogeneous and sequence-specific cohorts is warranted. (3) Different fat-suppression methods were used for the PROPELLER- and EPI-DWI sequences, and this difference may have affected both the qualitative and quantitative image-quality assessments, thus presenting a potential confounding factor in the comparison between the two sequences. (4) In the quantitative analysis, the ROIs were initially placed on the PROPELLER images and then copied to the conventional images. Although the ROI positions were adjusted when necessary using both image sets as references, this approach may still have introduced a mild bias toward the PROPELLER technique. (5) The imaging duration was not analyzed, and the PROPELLER technique tends to slightly prolong the scan time compared to conventional approaches [18]. However, given the marked improvements in contrast and artifact suppression demonstrated in head and neck imaging in this study, we propose that the use of the PROPELLER acquisition technique together with DWI offers significant diagnostic value and may warrant incorporation into routine protocols, even with a modest time penalty. Future research should focus on reducing the scan time while preserving the image quality in order to optimize the method's clinical feasibility.
5. Conclusion
The integration of DLR with the PROPELLER acquisition scheme significantly improved the image quality for both Fs-T2WI and DWI sequences of the head and neck compared to standard approaches. Accordingly, this technique may provide a reliable imaging approach for clinical assessments of head and neck lesions.
CRediT authorship contribution statement
Yukie Shimizu: Writing – review & editing, Investigation. Motoma Kanaya: Writing – review & editing, Investigation. Satoshi Kano: Writing – review & editing, Resources. Taisuke Harada: Writing – review & editing. Hiroyuki Kameda: Writing – review & editing. Yohei Ikebe: Writing – review & editing. Yuki Takano: Writing – original draft, Visualization, Validation, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Kohsuke Kudo: Writing – review & editing, Supervision, Project administration, Conceptualization. Akihiro Homma: Writing – review & editing, Resources. Noriyuki Fujima: Writing – review & editing, Validation, Supervision, Resources, Project administration, Methodology, Funding acquisition, Conceptualization.
Consent for publication
Not applicable.
Ethical statement
Our institutional review board approved this retrospective study (ID: 025–0362), and the requirement for written informed consent was waived. All procedures involving human participants were conducted in accordance with the ethical standards of the institutional and national research committees and the 1964 Declaration of Helsinki and its later amendments or comparable ethical guidelines.
Funding statement
This work was supported by a grant from the Japan Society for the Promotion of Science (JSPS) KAKENHI, No. JP25K10979.
Declaration of Competing Interest
The authors declare that they have no competing interests.
Data availability
The datasets used and analyzed during this study are available from the corresponding author upon reasonable request.
References
- 1.Christe A., Waldherr C., Hallett R., Zbaeren P., Thoeny H. MR imaging of parotid tumors: Typical lesion characteristics in MR imaging improve discrimination between benign and malignant disease. AJNR Am. J. Neuroradiol. 2011;32:1202–1207. doi: 10.3174/ajnr.A2520. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Saat R., Mahmood G., Laulajainen-Hongisto A., Lempinen L., Aarnisalo A.A., Jero J., Markkola A. Comparison of MR imaging findings in paediatric and adult patients with acute mastoiditis and incidental intramastoid bright signal on T2-weighted images. Eur. Radiol. 2016;26:2632–2639. doi: 10.1007/s00330-015-4113-5. [DOI] [PubMed] [Google Scholar]
- 3.Li Z., Wang X., Jiang H., Qu X., Wang C., Chen X., Chong V.F., Zhang L., Xian J. Chronic invasive fungal rhinosinusitis vs sinonasal squamous cell carcinoma: The differentiating value of MRI. Eur. Radiol. 2020;30:4466–4474. doi: 10.1007/s00330-020-06838-1. [DOI] [PubMed] [Google Scholar]
- 4.Akutsu A., Horikoshi T., Yokota H., Wada T., Motoori K., Nasu K., Yamasaki K., Hanazawa T., Ikeda J.I., Uno T. MR imaging findings of carcinoma ex pleomorphic adenoma related to extracapsular invasion and prognosis. AJNR Am. J. Neuroradiol. 2022;43:1639–1645. doi: 10.3174/ajnr.A7656. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Touska P., Connor S.E.J. Recent advances in MRI of the head and neck, skull base and cranial nerves: New and evolving sequences, analyses and clinical applications. Br. J. Radiol. 2019;92 doi: 10.1259/bjr.20190513. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Bhatnagar P., Subesinghe M., Patel C., Prestwich R., Scarsbrook A.F. Functional imaging for radiation treatment planning, response assessment, and adaptive therapy in head and neck cancer. Radiographics. 2013;33:1909–1929. doi: 10.1148/rg.337125163. [DOI] [PubMed] [Google Scholar]
- 7.King A.D., Thoeny H.C. Functional MRI for the prediction of treatment response in head and neck squamous cell carcinoma: Potential and limitations. Cancer Imaging. 2016;16:23. doi: 10.1186/s40644-016-0080-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Bhatt N., Gupta N., Soni N., Hooda K., Sapire J.M., Kumar Y. Role of diffusion-weighted imaging in head and neck lesions: Pictorial review. Neuroradiol. J. 2017;30:356–369. doi: 10.1177/1971400917708582. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Payabvash S. Quantitative diffusion magnetic resonance imaging in head and neck tumors. Quant. Imaging Med. Surg. 2018;8:1052–1065. doi: 10.21037/qims.2018.10.14. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Boucher F., Liao E., Srinivasan A. Diffusion-weighted imaging of the head and neck (including temporal bone) Magn. Reson. Imaging Clin. N. Am. 2021;29:205–232. doi: 10.1016/j.mric.2021.01.005. [DOI] [PubMed] [Google Scholar]
- 11.Hirano Y., Fujima N., Ishizaka K., Aoike T., Nakagawa J., Yoneyama M., Kudo K. Utility of echo planar imaging with compressed sensing-sensitivity encoding (EPICS) for the evaluation of the head and neck region. Cureus. 2024;16 doi: 10.7759/cureus.54203. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Pipe J.G. Motion correction with PROPELLER MRI: Application to head motion and free-breathing cardiac imaging. Magn. Reson. Med. 1999;42:963–969. doi: 10.1002/(sici)1522-2594(199911)42:5<963::aid-mrm17>3.0.co;2-l. 〈10.1002/(sici)1522-2594(199911)42:5<963::aid-mrm17>3.0.co;2-l〉 [DOI] [PubMed] [Google Scholar]
- 13.Bayramoglu S., Kilickesmez O., Cimilli T., Kayhan A., Yirik G., Islim F., Alibek S. T2-weighted MRI of the upper abdomen: Comparison of four fat-suppressed T2-weighted sequences including PROPELLER (BLADE) technique. Acad. Radiol. 2010;17:368–374. doi: 10.1016/j.acra.2009.10.015. [DOI] [PubMed] [Google Scholar]
- 14.Lavdas E., Mavroidis P., Kostopoulos S., Glotsos D., Roka V., Topalzikis T., Bakas A., Oikonomou G., Papanikolaou N., Batsikas G., Kaffes I., Kechagias D. Improvement of image quality using BLADE sequences in brain MR imaging. Magn. Reson. Imaging. 2013;31:189–200. doi: 10.1016/j.mri.2012.08.001. [DOI] [PubMed] [Google Scholar]
- 15.Liu Y.J., Lee Y.H., Chang H.C., Chung H.W., Wang C.W., Juan C.H., Chu Y.H., Lee J.C., Juan C.J. Imaging quality of PROPELLER diffusion-weighted MR imaging and its diagnostic performance in distinguishing pleomorphic adenomas from Warthin tumors of the parotid gland. NMR Biomed. 2020;33 doi: 10.1002/nbm.4282. [DOI] [PubMed] [Google Scholar]
- 16.Li W., Shi J., Bian W., Li J., Chen X., Feng J., Yu J., Wang J., Niu J. Performance of PROPELLER FSE T2WI in reducing metal artifacts of material porcelain fused to metal crown: A clinical preliminary study. Sci. Rep. 2022;12:8442. doi: 10.1038/s41598-022-12402-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Mavroidis P., Giankou E., Papaioannou M., Roka V., Tsikrika A., Kostopoulos S., Glotsos D., Sakkas G.K., Dardiotis E., Chaniotis D., Kapsalaki E., Lavdas E. Comparison between EPI DWI and PROPELLER DWI in brain MR imaging. Curr. Probl. Diagn. Radiol. 2024;53:73–80. doi: 10.1067/j.cpradiol.2023.08.015. [DOI] [PubMed] [Google Scholar]
- 18.Yamauchi H., Srinivasan A. Diffusion imaging of the head and neck. Curr. Radiol. Rep. 2014;2:49. doi: 10.1007/s40134-014-0049-9. [DOI] [Google Scholar]
- 19.Fujima N., Kamagata K., Ueda D., Fujita S., Fushimi Y., Yanagawa M., Ito R., Tsuboyama T., Kawamura M., Nakaura T., Yamada A., Nozaki T., Fujioka T., Matsui Y., Hirata K., Tatsugami F., Naganawa S. Current state of artificial intelligence in clinical applications for head and neck MR imaging. Magn. Reson. Med. Sci. 2023;22:401–414. doi: 10.2463/mrms.rev.2023-0047. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Nishioka N., Fujima N., Tsuneta S., Yoshikawa M., Kimura R., Sakamoto K., Kato F., Miyata H., Kikuchi H., Matsumoto R., Abe T., Kwon J., Yoneyama M., Kudo K. Enhancing the image quality of prostate diffusion-weighted imaging in patients with prostate cancer through model-based deep learning reconstruction. Eur. J. Radiol. Open. 2024;13 doi: 10.1016/j.ejro.2024.100588. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Nagaraj U.D., Dillman J.R., Tkach J.A., Greer J.S., Leach J.L. Evaluation of 3D T1-weighted spoiled gradient echo MR image quality using artificial intelligence image reconstruction techniques in the pediatric brain. Neuroradiology. 2024;66:1849–1857. doi: 10.1007/s00234-024-03417-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Fujima N., Nakagawa J., Kameda H., Ikebe Y., Harada T., Shimizu Y., Tsushima N., Kano S., Homma A., Kwon J., Yoneyama M., Kudo K. Improvement of image quality in diffusion-weighted imaging with model-based deep learning reconstruction for evaluations of the head and neck. MAGMA. 2024;37:439–447. doi: 10.1007/s10334-023-01129-4. [DOI] [PubMed] [Google Scholar]
- 23.Ikeda H., Ohno Y., Yamamoto K., Murayama K., Ikedo M., Yui M., Kumazawa Y., Shimamura Y., Takagi Y., Nakagaki Y., Hanamatsu S., Obama Y., Ueda T., Nagata H., Ozawa Y., Iwase A., Toyama H. Deep learning reconstruction for DWIs by EPI and FASE sequences for head and neck tumors. Cancers (Basel) 2024;16:1714. doi: 10.3390/cancers16091714. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Nitschke L.V., Lerchbaumer M., Ulas T., Deppe D., Nickel D., Geisel D., Kubicka F., Wagner M., Walter-Rittel T. Deep learning CAIPIRINHA-VIBE improves and accelerates head and neck MRI. Acad. Radiol. 2025;32:5507–5515. doi: 10.1016/j.acra.2025.05.020. [DOI] [PubMed] [Google Scholar]
- 25.Fujima N., Shimizu Y., Ikebe Y., Kameda H., Harada T., Tsushima N., Kano S., Homma A., Kwon J., Yoneyama M., Kudo K. Dual-type deep learning-based image reconstruction for advanced denoising and super-resolution processing in head and neck T2-weighted imaging. Jpn. J. Radiol. 2025;43:1097–1105. doi: 10.1007/s11604-025-01756-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Lebel R.M. Perform. Charact. a Nov. Deep. Learn. -Based MR Image Reconstr. Pipeline arXiv 2008. 06559. 2020 doi: 10.48550/arXiv.2008.06559. [DOI] [Google Scholar]
- 27.Fujima N., Nakagawa J., Ikebe Y., Kameda H., Harada T., Shimizu Y., Tsushima N., Kano S., Homma A., Kwon J., Yoneyama M., Kudo K. Improved image quality in contrast-enhanced 3D-T1 weighted sequence by compressed sensing-based deep-learning reconstruction for the evaluation of head and neck. Magn. Reson. Imaging. 2024;108:111–115. doi: 10.1016/j.mri.2024.02.006. [DOI] [PubMed] [Google Scholar]
- 28.Kaniewska M., Deininger-Czermak E., Getzmann J.M., Wang X., Lohezic M., Guggenberger R. Application of deep learning-based image reconstruction in MR imaging of the shoulder joint to improve image quality and reduce scan time. Eur. Radiol. 2023;33:1513–1525. doi: 10.1007/s00330-022-09151-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Hahn S., Yi J., Lee H.J., Lee Y., Lee J., Wang X., Fung M. Comparison of deep learning-based reconstruction of PROPELLER shoulder MRI with conventional reconstruction. Skelet. Radiol. 2023;52:1545–1555. doi: 10.1007/s00256-023-04321-8. [DOI] [PubMed] [Google Scholar]
- 30.Hirokawa Y., Isoda H., Maetani Y.S., Arizono S., Shimada K., Togashi K. MRI artifact reduction and quality improvement in the upper abdomen with PROPELLER and prospective acquisition correction (PACE) technique. AJR Am. J. Roentgenol. 2008;191:1154–1158. doi: 10.2214/AJR.07.3657. [DOI] [PubMed] [Google Scholar]
- 31.Ciet P., Serra G., Bertolo S., Spronk S., Ros M., Fraioli F., Quattrucci S., Assael M.B., Catalano C., Pomerri F., Tiddens H.A., Morana G. Assessment of CF lung disease using motion corrected PROPELLER MRI: A comparison with CT. Eur. Radiol. 2016;26:780–787. doi: 10.1007/s00330-015-3850-9. [DOI] [PubMed] [Google Scholar]
- 32.Forbes K.P., Pipe J.G., Bird C.R., Heiserman J.E. PROPELLER MRI: Clinical testing of a novel technique for quantification and compensation of head motion. J. Magn. Reson. Imaging. 2001;14:215–222. doi: 10.1002/jmri.1176. [DOI] [PubMed] [Google Scholar]
- 33.Arfanakis K., Tamhane A.A., Pipe J.G., Anastasio M.A. k-space undersampling in PROPELLER imaging. Magn. Reson. Med. 2005;53:675–683. doi: 10.1002/mrm.20380. [DOI] [PubMed] [Google Scholar]
- 34.Song Y.S., Lee I.S., Hwang M., Jang K., Wang X., Fung M. Clinical efficacy of motion-insensitive imaging technique with deep learning reconstruction to improve image quality in cervical spine MR imaging. Br. J. Radiol. 2024;97:812–819. doi: 10.1093/bjr/tqae026. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Pocepcova V., Zellner M., Callaghan F., Wang X., Lohezic M., Geiger J., Kellenberger C.J. Deep learning-based denoising image reconstruction of body magnetic resonance imaging in children. Pediatr. Radiol. 2025;55:1235–1244. doi: 10.1007/s00247-025-06230-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Farahani K., Sinha U., Sinha S., Chiu L.C., Lufkin R.B. Effect of field strength on susceptibility artifacts in magnetic resonance imaging. Comput. Med. Imaging Graph. 1990;14:409–413. doi: 10.1016/0895-6111(90)90040-i. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
The datasets used and analyzed during this study are available from the corresponding author upon reasonable request.